Splinter is an open source tool for testing web applications using Python. It lets you automate browser actions, such as visiting URLs and interacting with their items.
http://splinter.cobrateam.info/
It is a clone of scrapy_model but intead of Scrapy it uses Splinter as engine, so it allows scraping of JavaScript websites.
Everything is in to-do, so just dont use it until it is released to PyPI.
This module should keep implement the same api as scrapy_model, at least the support for CSSField, XPathField, processors, validators, multiple queries and parse_methods.
Also it should implement a layer to interact with JavaScript.
It is just a helper to create scrapers using the Scrapy Selectors allowing you to select elements by CSS or by XPATH and structuring your scraper via Models (just like an ORM model) and plugable to an ORM model via populate method.
Import the BaseFetcherModel, CSSField or XPathField (you can use both)
from splinter_model import BaseFetcherModel, CSSFieldGo to a webpage you want to scrap and use chrome dev tools or firebug to figure out the css paths then considering you want to get the following fragment from some page.
<span id="person">Bruno Rocha <a href="http://brunorocha.org">website</a></span>class MyFetcher(BaseFetcherModel):
name = CSSField('span#person')
website = CSSField('span#person a')
# XPathField('//xpath_selector_here')Fields can receive auto_extract=True parameter which auto extracts values from selector before calling the parse or processors. Also you can pass the takes_first=True which will for auto_extract and also tries to get the first element of the result, because scrapy selectors returns a list of matched elements.
You can use multiple queries for a single field
name = XPathField(
['//*[@id="8"]/div[2]/div/div[2]/div[2]/ul',
'//*[@id="8"]/div[2]/div/div[3]/div[2]/ul']
)In that case, the parsing will try to fetch by the first query and returns if finds a match, else it will try the subsequent queries until it finds something, or it will return an empty selector.
If you want to run multiple queries and also validates the best match you can pass a validator function which will take the scrapy selector an should return a boolean.
Example, imagine you get the "name" field defined above and you want to validates each query to ensure it has a 'li' with a text "Schblaums" in there.
def has_schblaums(selector):
for li in selector.css('li'): # takes each <li> inside the ul selector
li_text = li.css('::text').extract() # Extract only the text
if "Schblaums" in li_text: # check if "Schblaums" is there
return True # this selector is valid!
return False # invalid query, take the next or default value
class Fetcher(....):
name = XPathField(
['//*[@id="8"]/div[2]/div/div[2]/div[2]/ul',
'//*[@id="8"]/div[2]/div/div[3]/div[2]/ul'],
query_validator=has_schblaums,
default="undefined_name" # optional
)In the above example if both queries are invalid, the "name" field will be filled with an empty_selector, or the value defined in "default" parameter.
NOTE: if the field has a "default" and fails in all the matcher, the default value will be passed to "processor" and also to "parse_" methods.
Every method named parse_<field> will run after all the fields are fetched for each field.
def parse_name(self, selector):
# here selector is the scrapy selector for 'span#person'
name = selector.css('::text').extract()
return name
def parse_website(self, selector):
# here selector is the scrapy selector for 'span#person a'
website_url = selector.css('::attr(href)').extract()
return website_urlafter defined need to run the scraper
fetcher = Myfetcher(url='http://.....') # optionally you can use cached_fetch=True to cache requests on redis
fetcher.parse()Now you can iterate _data, _raw_data and atributes in fetcher
>>> fetcher.name
<CSSField - name - Bruno Rocha>
>>> fetcher.name.value
Bruno Rocha
>>> fetcher._data
{"name": "Bruno Rocha", "website": "http://brunorocha.org"}You can populate some object
>>> obj = MyObject()
>>> fetcher.populate(obj) # fields optional
>>> obj.name
Bruno RochaIf you do not want to define each field explicitly in the class, you can use a json file to automate the process
class MyFetcher(BaseFetcherModel):
""" will load from json """
fetcher = MyFetcher(url='http://.....')
fetcher.load_mappings_from_file('path/to/file.json')
fetcher.parse()In that case file.json should be
{
"name": {"css", "span#person"},
"website": {"css": "span#person a"}
}You can use {"xpath": "..."} in case you prefer select by xpath
There are 2 ways of transforming or normalizing the data for each field
A processor is a function, or a list of functions which will be called in the given sequence against the field value, it receives the raw_selector or the value depending on auto_extract and takes_first arguments.
It can be used for Normalization, Clean, Transformation etc..
Example:
def normalize_state(state_name):
# query my database and return the first instance of state object
return MyDatabase.State.Search(name=state_name).first()
def text_cleanup(state_name):
return state_name.strip().replace('-', '').lower()
class MyFetcher(BaseFetcherModel):
state = CSSField(
"#state::text",
takes_first=True,
processor=[text_cleanup, normalize_state]
)
fetcher = MyFetcher(url="http://....")
fetcher.parse()
fetcher._raw_data.state
'Sao-Paulo'
fetcher._data.state
<ORM Instance - State - São Paulo>any method called parse_<field_name> will run after all the process of selecting and parsing, it receives the selector or the value depending on auto_extract and takes_first argument in that field.
example:
def parse_name(self, selector):
return selector.css('::text').extract()[0].upper()In the above case, the name field returns the raw_selector and in the parse method we can build extra queries using css or xpath and also we need to extract() the values from the selector and optionally select the first element and apply any transformation we need.
In order to cache the html returned by the url fetching for future parsing and tests you specify a cache model, by default there is no cache but you can use the built in RedisCache passing
from splinter_model import RedisCache
fetcher = TestFetcher(cache_fetch=True,
cache=RedisCache,
cache_expire=1800)or specifying arguments to the Redis client.
it is a general Redis connection from python
redismodule
fetcher = TestFetcher(cache_fetch=True,
cache=RedisCache("192.168.0.12:9200"),
cache_expire=1800)You can create your own caching structure, e.g: to cache htmls in memcached or s3
the cache class just need to implement get and set methods.
from boto import connect_s3
class S3Cache(object):
def __init__(self, *args, **kwargs):
connection = connect_s3(ACCESS_KEY, SECRET_KEY)
self.bucket = connection.get_bucket(BUCKET_ID)
def get(self, key):
value = self.bucket.get_key(key)
return value.get_contents_as_string() if key else None
def set(self, key, value, expire=None):
self.bucket.set_contents(key, value, expire=expire)
fetcher = MyFetcher(url="http://...",
cache_fetch=True,
cache=S3cache,
cache_expire=1800)easy to install
If running ubuntu maybe you need to run:
sudo apt-get install python-scrapy
sudo apt-get install libffi-dev
sudo apt-get install python-devthen
pip install splinter_modelor
git clone https://github.com/rochacbruno/splinter_model
cd splinter_model
pip install -r requirements.txt
python setup.py install
python example.pyExample code to fetch the url http://en.m.wikipedia.org/wiki/Guido_van_Rossum
#coding: utf-8
from splinter_model import BaseFetcherModel, CSSField, XPathField
class TestFetcher(BaseFetcherModel):
photo_url = XPathField('//*[@id="content"]/div[1]/table/tr[2]/td/a')
nationality = CSSField(
'#content > div:nth-child(1) > table > tr:nth-child(4) > td > a',
)
links = CSSField(
'#content > div:nth-child(11) > ul > li > a.external::attr(href)',
auto_extract=True
)
def parse_photo_url(self, selector):
return "http://en.m.wikipedia.org/{}".format(
selector.xpath("@href").extract()[0]
)
def parse_nationality(self, selector):
return selector.css("::text").extract()[0]
def parse_name(self, selector):
return selector.extract()[0]
def pre_parse(self, selector=None):
# this method is executed before the parsing
# you can override it, take a look at the doc string
def post_parse(self):
# executed after all parsers
# you can load any data on to self._data
# access self._data and self._fields for current data
# self.selector contains original page
# self.fetch() returns original html
self._data.url = self.url
class DummyModel(object):
"""
For tests only, it can be a model in your database ORM
"""
if __name__ == "__main__":
from pprint import pprint
fetcher = TestFetcher(cache_fetch=True)
fetcher.url = "http://en.m.wikipedia.org/wiki/Guido_van_Rossum"
# Mappings can be loaded from a json file
# fetcher.load_mappings_from_file('path/to/file')
fetcher.mappings['name'] = {
"css": ("#section_0::text")
}
fetcher.parse()
print "Fetcher holds the data"
print fetcher._data.name
print fetcher._data
# How to populate an object
print "Populating an object"
dummy = DummyModel()
fetcher.populate(dummy, fields=["name", "nationality"])
# fields attr is optional
print dummy.nationality
pprint(dummy.__dict__)Fetcher holds the data
Guido van Rossum
{'links': [u'http://www.python.org/~guido/',
u'http://neopythonic.blogspot.com/',
u'http://www.artima.com/weblogs/index.jsp?blogger=guido',
u'http://python-history.blogspot.com/',
u'http://www.python.org/doc/essays/cp4e.html',
u'http://www.twit.tv/floss11',
u'http://www.computerworld.com.au/index.php/id;66665771',
u'http://www.stanford.edu/class/ee380/Abstracts/081105.html',
u'http://stanford-online.stanford.edu/courses/ee380/081105-ee380-300.asx'],
'name': u'Guido van Rossum',
'nationality': u'Dutch',
'photo_url': 'http://en.m.wikipedia.org//wiki/File:Guido_van_Rossum_OSCON_2006.jpg',
'url': 'http://en.m.wikipedia.org/wiki/Guido_van_Rossum'}
Populating an object
Dutch
{'name': u'Guido van Rossum', 'nationality': u'Dutch'}